1. Genomics and Evolutionary Biology
Download icon

Genetic variation in offspring indirectly influences the quality of maternal behaviour in mice

  1. David George Ashbrook Is a corresponding author
  2. Beatrice Gini
  3. Reinmar Hager Is a corresponding author
  1. University of Manchester, United Kingdom
Research Article
Cited
3
Views
1,281
Comments
0
Cite as: eLife 2015;4:e11814 doi: 10.7554/eLife.11814

Abstract

Conflict over parental investment between parent and offspring is predicted to lead to selection on genes expressed in offspring for traits influencing maternal investment, and on parentally expressed genes affecting offspring behaviour. However, the specific genetic variants that indirectly modify maternal or offspring behaviour remain largely unknown. Using a cross-fostered population of mice, we map maternal behaviour in genetically uniform mothers as a function of genetic variation in offspring and identify loci on offspring chromosomes 5 and 7 that modify maternal behaviour. Conversely, we found that genetic variation among mothers influences offspring development, independent of offspring genotype. Offspring solicitation and maternal behaviour show signs of coadaptation as they are negatively correlated between mothers and their biological offspring, which may be linked to costs of increased solicitation on growth found in our study. Overall, our results show levels of parental provisioning and offspring solicitation are unique to specific genotypes.

https://doi.org/10.7554/eLife.11814.001

eLife digest

Genes encode instructions that can influence the behaviour and physical traits of the individual that carries them. Individuals of the same species can carry different versions (or variants) of the same gene, leading to a variety of traits in the population. However, a gene expressed in one individual can also alter the traits of another individual. This is known as an indirect genetic effect. For example, a gene in a mother that affects her ability to provide care may influence how her offspring develop.

Researchers have predicted that offspring should be able to manipulate their mothers to try and gain more care than the mothers are willing to give. Furthermore, the offspring born to mothers who respond to this begging are predicted to save energy and beg less. However, few gene variants that indirectly modify the behaviour of mothers or offspring have so far been identified.

Ashbrook et al. used mice to test the idea that genetic variation in particular locations in the offspring’s genome can affect maternal behaviour, and vice versa. In the first experiment, mother mice with different gene variants fostered litters of mouse pups that were all genetically identical. In the other experiment, genetically identical mothers fostered litters of mouse pups with different gene variants. Ashbrook et al. identified several locations in the offspring’s genome that modified the behaviour of their foster mothers. Furthermore, the experiments also show that genetic variation among the mothers influenced the development of their offspring, independent of the genes carried by the offspring.

The next steps are to identify the specific genes underlying the changes in behaviour, and the molecular and genetic pathways by which they impact indirectly on the traits of other individuals.

https://doi.org/10.7554/eLife.11814.002

Introduction

The close interaction between mother and offspring in mammals is fundamental to offspring development and fitness. However, parent and offspring are in conflict over how much parents should invest in their young where offspring typically demand more than is optimal for the parent (Trivers, 1974; Godfray, 1995), and the existence of this genetic conflict has been demonstrated in empirical research (Kölliker et al., 2015). The resulting selection pressures are predicted to lead to the evolution of traits in offspring that influence parental behaviour (and thus investment). Conversely, parental traits should be selected for their effects on offspring traits that influence parental behaviour indirectly (Kilner and Hinde, 2012). The correlation between parental and offspring traits has been the focus of coadaptation models where specific combinations of demand and provisioning are selectively favoured (Wolf and Brodie III, 1998; Kölliker et al., 2005). The fundamental assumption underlying predictions about the evolution of traits involved in parent-offspring interactions is that genetic variation in offspring exists for traits that indirectly influence maternal investment and vice versa. However, it remains to be shown whether specific genetic variants in offspring indirectly influence maternal behaviour. In an experimental mouse population, we demonstrate that genes expressed in offspring modify the quality of maternal behaviour and thus affect, indirectly, offspring fitness.

To investigate the genetics of parent-offspring interactions we conducted a cross-fostering experiment between genetically variable and genetically uniform mice, using the largest genetic reference panel in mammals, the BXD mouse population. We generated families of genetically variable mothers and genetically uniform offspring by cross-fostering C57BL/6J (B6) litters, in which no genetic variation occurs between animals of this strain, to mothers of a given BXD strain. Conversely, a BXD female’s litter was cross-fostered to B6 mothers (Figure 1). Thus, we can analyse the effects of genetic variation in mothers or offspring while controlling for genetic variation in the other. This cross-fostering design has been successfully utilized in previous studies on family interactions because it breaks the correlation between maternal and offspring traits. Here, different families, or naturally occurring variation of maternal and offspring trait combinations across different broods, are assumed to represent distinct evolved strategies (Agrawal et al., 2001; Hager and Johnstone, 2003; Meunier and Kölliker, 2012). From birth until weaning at 3 weeks of age we recorded offspring and maternal body weights and behaviour, following Hager and Johnstone (2003).

Experimental cross-foster design.

Females of different lines of the BXD strain (light to grey mice) adopt B6 offspring (dark) and B6 females (dark) adopt offspring born to females of different BXD lines. A total of 42 BXD lines with three within-line repeats plus the corresponding B6 families were set up for the experiment.

https://doi.org/10.7554/eLife.11814.003

Results

We first investigated whether there is evidence for indirect genetic effects in offspring influencing maternal behaviour, keeping maternal genotype constant. To find out if maternal behaviour is modified by genes expressed in offspring, we mapped variation in maternal behaviour as a function of their adoptive BXD offspring genotype. We found that variation in offspring genotype affects maternal behaviour, which in turn influences offspring development and fitness. Throughout, we denote loci as either maternal or offspring, Mat or Osp, followed by whether it is an indirect effect (Ige) or a direct effect locus (Dge). We provide a summary of all loci in Table 1, and mapping details in Supplementary files 1A-D. During the first postnatal week we mapped a locus on offspring chromosome 7, OspIge7.1, modifying maternal nestbuidling on day 6 (Figure 2). At this locus, the D2 allele increases the trait value such that B6 mothers showed more nestbuilding activity when fostering BXD pups carrying the D2 allele at the locus. Nestbuilding is particularly important for offspring fitness as thermoregulation is underdeveloped and hypothermia is the primary cause of early death, even if milk is supplied (Lynch and Possidente, 1978). We detected a further locus on offspring distal chromosome 5 (OspIge5.1) that affects maternal behaviour on day 14, around the time we expect the weaning conflict to be highest (Figure 3). Here, mothers showed increased levels of maternal behaviour when fostering BXD offspring carrying the D2 allele. Conversely, we can look at how variation in maternal genotype influences offspring traits, keeping offspring genotype constant, mapping variation in offspring traits as a function of BXD genotype. Here, we found that offspring growth during the second week is affected by a locus on maternal chromosome 17, MatIge17.1, where the B6 allele increases the trait value (Figure 4). In addition to looking at indirect genetic effects we can analyse direct genetic effects, i.e. how an individual’s genotype influences its own traits. In mothers we found a direct genetic effect locus for maternal behaviour on proximal chromosome 10, and for nestbuilding behaviour on chromosome 1 (MatDge10.1 and MatDge1.1, respectively), where the B6 allele increases the trait value in both cases. We also detected a locus for offspring solicitation behaviour on chromosome 5 (OspDge5.1) with the D2 allele increasing the level of solicitation shown. Its location is at the opposite end on chromosome 5 from where OspIge5.1 is located.

Offspring indirect genetic effect modifying maternal nestbuidling behaviour (OspIge7.1).

The figure shows an offspring genomescan of maternal nestbuilding behaviour on day 6. The blue line represents the genome scan, showing the likelihood ratio statistic (LRS) associated with each marker across the 19 autosomal and the X chromosome. The top, pink, line marks genome-wide significance, the lower, grey, line the suggestive significance threshold. The green or red line show the additive coefficient, with green showing that the DBA/2J alleles increase trait values and red that the C57BL/6J alleles increase trait values. The green axis on the right shows by how much the respective alleles increase trait values (the DBA/2J allele in offspring increases maternal nestbuilding by ~0.8).

https://doi.org/10.7554/eLife.11814.004
Offspring indirect genetic effect modifying maternal behaviour (OspIge5.1).

The figure shows an offspring genomescan of maternal behaviour on day 14. The blue line represents the genome scan, showing the likelihood ratio statistic (LRS) associated with each marker across the 19 autosomal and the X chromosome. The top, pink, line marks genome-wide significance, the lower, grey, line the suggestive significance threshold. The green or red line show the additive coefficient, with green showing that the DBA/2J alleles increase trait values and red that the C57BL/6J alleles increase trait values. The green axis on the right shows by how much the respective alleles increase trait values (the DBA/2J allele in offspring increases maternal behaviour by ~2.5).

https://doi.org/10.7554/eLife.11814.006
Maternal indirect genetic effect modifying offspring growth (MatIge17).

The figure shows a maternal genomescan of offspring growth in the second postnatal week. The blue line represents the genome scan, showing the likelihood ratio statistic (LRS) associated with each marker across the 19 autosomal and the X chromosome. The top, pink, line marks genome-wide significance, the lower, grey, line the suggestive significance threshold. The green or red line show the additive coefficient, with green showing that the DBA/2J alleles increase trait values and red that the C57BL/6J alleles increase trait values. The green axis on the right shows by how much the respective alleles increase trait values (the C57BL/6J allele in mothers increases offspring growth by ~0.25).

https://doi.org/10.7554/eLife.11814.008
Table 1

Summary of direct and indirect genetic effects. We list the loci, followed by their position in Mb, the 1.5 LOD confidence interval (Dupuis and Siegmund, 1999), the genome-wide peak marker LRS and LOD score and associated p-value and the number of genes within the interval.

https://doi.org/10.7554/eLife.11814.010
LociQTL positionConfidence intervalMax LRSMax LODMax PNumber of genes
MatDge1.1 168.32165.24 - 172.0617.033.700.06968
OspDge5.1 23.82717.82 - 24.6217.853.880.04673
OspIge5.1 146.68145.20 - 147.6518.654.050.03840
OspIge7.1 53.6847.76 - 56.6517.853.880.039232
81.4976.12 - 90.9216.063.490.087133
MatDge10.1 19.0918.61 - 21.8322.374.860.00830
MatIge17.1 23.3211.48 - 31.1719.024.130.022422
33.0231.32 - 40.6518.574.030.028321

Overall, our results show that IGEs, as well as DGEs, can be linked to specific loci that affect parent-offspring interaction, and it is thus possible that selection may occur on genes with indirect and direct effects on parental behaviour. Importantly, our result that variation in maternal behaviour is affected by genes expressed in offspring is also clearly borne out at the phenotypic level: offspring solicitation behaviour in genetically variable BXD pups is positively correlated with the level of maternal behaviour in their genetically uniform B6 adoptive mothers on all three days we measured behaviour with significant effects of maternal behaviour and day on offspring solicitation (GLM, F1,103 = 17.62, P < 0.001 and F2,103 = 14.22, P < 0.001, respectively; day 6: Pearsons’r = 0.56, P = 0.003; day 10: r = 0.63, P = 0.001, and day 14: r = 0.55, P = 0.008; Supplementary file 2, a). Similarly, we can investigate how traits in genetically uniform B6 offspring correlate with maternal traits in their genetically variable BXD mothers (i.e. within adoptive families). We found that maternal behaviour and day have a significant positive effect on offspring solicitation behaviour (GLM, F1,118 = 6.33, P = 0.013, and F2,118 = 6.33, P < 0.001; Supplementary file 2, b). While one might generally assume that mothers behave in response to offspring solicitation behaviour, these results show, perhaps surprisingly, that variation in maternal behaviour influences the level of solicitation: here, we need to remember that there is no variation in offspring genotype so we assume that across families differences in offspring behaviour are due to differences in the genotype of their adoptive mothers.

Coadaptation of parental and offspring traits

While in the previous section we have focused on analysing traits within foster families, we now turn to the correlation between traits of biological families, i.e. mothers and their biological offspring. This correlation has been analysed in coadaptation models, which make specific predictions about how parental and offspring traits are correlated, and in empirical work (Kölliker et al., 2000; Agrawal et al., 2001; Curley et al., 2004; Lock et al., 2004; Kölliker et al., 2005; Hinde et al., 2010; Meunier and Kölliker, 2012). Prior experimental studies have found both a positive (Parus major; Kölliker et al., 2000: Nicrophorus vespilloides; Lock et al., 2004) and negative correlation between offspring solicitation and parental traits (Sehirus cinctus; Agrawal et al., 2001). In our study, we found a negative correlation. When we measured short-term provisioning, we found a negative correlation between BXD offspring short-term weight gain and the corresponding provisioning of their biological (BXD) mothers on day 10, as well as a negative correlation between BXD offspring solicitation and the corresponding provisioning of their biological (BXD) mothers on day 14 (GLM, F1,32 = 4.77, P = 0.036; r = -0.34 and GLM, F1,28 = 8.046, P = 0.008, r = -0.48; Figure 5 and Supplementary file 2, d). Thus, our results suggest that mothers who are generous providers produce young that solicit less maternal resources than offspring born to less generous mothers. Such a negative correlation is predicted to occur when maternal traits are predominantly under selection as long as parents respond to offspring demand (which we have shown above; Kölliker et al., 2005). One scenario to explain this negative correlation might be that each BXD line, i.e. genotype, is characterized by a unique (to this line, everything else being equal) combination of offspring and maternal behaviours where higher maternal provisioning is correlated with lower offspring solicitation. This may be due to the cost of increased solicitation (reflected in reduced bodyweight for the effort expended) for which we found evidence in our study. Bodyweight is indeed negatively correlated with the level of offspring solicitation (GLM, F1,66 = 20.57, P < 0.001 e.g. day 10, r = -0.39, and day 14, r = -0.44; Figure 6 and Supplementary file 2, e).

Correlation between offspring and maternal traits in biological BXD families.

The first panel shows the correlation between BXD offspring short-term weight change per pup and provisioning of their corresponding biological BXD mother on day 10 per pup. The second panel shows the correlation between the level of BXD offspring solicitation per pup on day 14 and their mother’s provisioning per pup.

https://doi.org/10.7554/eLife.11814.011
Correlation between per pup offspring solicitation and corresponding body weight in BXD lines on day 10 and day 14, respectively.
https://doi.org/10.7554/eLife.11814.013

Discussion

Our study of the genetics underlying family interactions has revealed that genes expressed in offspring can indirectly influence the quality of maternal behaviour and thus offspring fitness. At the same time, we detected specific loci in maternal genotype that indirectly modify offspring traits, which shows that IGEs can be an important component of the genetic architecture of complex traits (Bijma and Wade, 2008). We note that while postnatal cross-fostering controls for postnatal maternal effects, prenatal maternal effects can only be addressed (to some degree) by embryo transfer (e.g. Cowley et al., 1989), a procedure that is impractical in genetics experiments. Potentially, this may strengthen or weaken, for example, effects of BXD genotype on B6 maternal phenotype. At the same time, pre-natal maternal effects may also contribute to a phenotypic correlation between parent and biological offspring traits, as reported above on coadaptation. In-utero effect variation due to differences in BXD genotype may therefore contribute to differences in BXD offspring behaviour, in turn affecting the behaviour of their adoptive mothers. We now need to investigate the candidates identified here and how their effects on parental and offspring traits are integrated into the gene networks determining individual development. By controlling for genetic variation in either mothers or offspring we have been able to show that levels of maternal provisioning and offspring solicitation are unique to specific genotypes (here each BXD line) and that solicitation is costly. The ability to conduct complex systems genetics analyses in experimental systems of parent offspring interactions will enable us to concentrate now on understanding the underlying pathways involved, and how they are modified by social environmental conditions that determine adult phenotypes and associated reproductive success.

Materials and methods

Animals and genetic analysis

We used mice of the BXD recombinant inbred population, which consists of experimentally tractable and genetically defined mouse lines capturing a large amount of naturally occurring genetic variation, which underlies variation at the phenotypic level (e.g. Chesler et al., 2005; Hayes et al., 2014). The BXD panel incorporates ~5 million segregating SNPs, 500,000 insertions and deletions, and 55,000 copy-number variants. These lines are used for complex systems genetics analyses integrating massive phenotype and gene expression data sets obtained across years and studies (e.g. Andreux et al., 2012; Ashbrook et al., 2014).

The 42 BXD strains used in this study (1, 11, 12, 14, 24, 32, 34, 38–40, 43–45, 48a, 49–51, 55, 56, 60–64, 66–71, 73a, 73b, 73–75, 83, 84, 87, 89, 90, 98, 102) were obtained from Professor Robert W. Williams at the University of Tennessee Health Science Centre, Memphis, TN. C57BL/6J (B6) mice were obtained from Charles River, UK. Three within-line repeats plus the corresponding 42 B6 families with three within-line repeats were set up for the experiment (Figure 1). Sample size was determined considering power analyses and logistical aspects. Mapping power is maximised with increasing number of lines whereas within-line repeats n increase confidence of line average phenotypes, which, however, rapidly diminishes as n exceeds four (Belknap, 1998). We have modelled power and effect sizes following (Belknap, 1998): n = (Zα + Zβ)2 / S2QTL / S2Res). Zα and Zβ are Z values for a given α and β; S2QTL is the phenotypic variance due to a QTL and S2Res is the residual variance. With power (1-β) of 80%, α of 0.05 we estimated that with 45 lines we can detect QTL at genome-wide significance explaining ~ 16% of trait variance, which is sufficient mapping power given effect sizes of prior work. We have modelled the relationship between power and number of replicates using qtlDesign (Sen et al., 2007). Everything else being equal power can be optimized by maximizing the number of genotypes (i.e. lines) and reducing replicates, even with varying degrees of heritability. Thus, we set up three replicates using 45 lines, and line averages are mapped, although for some traits and some lines this number may be lower due to lower breeding success. Outliers have been retained as they represent distinct genotypes, evinced by outliers for different traits being from the same line. Interval mapping (Haley and Knott, 1992) relies on 3795 informative SNP markers across all chromosomes, except Y, as implemented in GeneNetwork (GN) (Hager et al., 2012). The BXD strains were genotyped using the MUGA array in 2011, along with genotypes generated earlier using Affymetrix and Illumina platforms (Shifman et al., 2006), and mm9 is used. Loci are identified in GN by the computation of a likelihood statistic score and significance was determined using 2000 permutations of the phenotype data. Candidates were identified within the region defined by using GeneNetwork (http://www.genenetwork.org) and further information combined from QTLminer (Alberts and Schughart, 2010), Entrez genes (http://www.ncbi.nlm.nih.gov/gene) and Mouse Genome Informatics (Eppig et al., 2015).

Behavioural protocols

Mice were maintained under standard laboratory conditions in the same room, exclusively used for the experiment in individually ventilated cages (IVC Tecniplast Green line), and given chow and water ad lib. Humidity ranges between 50% and 65% relative humidity, temperature between 20°C and 21°C. All animals were kept on a reverse dark light cycle with 12h red light (active phase) and 12h white light.

Cross-fostering of entire litters took place within 24h of birth of corresponding B6 and BXD females and analyses used trait values per pup to adjust for differences in adoptive litter size, which is a significant covariate for provisioning and solicitation as of course a mother nursing a large litter will overall provide more than a mother nursing a small litter. Both mothers and litters were weighed at birth and once weekly, for three weeks until weaning, i.e. at the end of week 1, week 2 and week 3, respectively, to enable the calculation of growth during these periods. In addition, we recorded maternal and offspring behaviour on postnatal days 6, 10 and 14 when we simulated maternal departure to standardize observation conditions (Hager and Johnstone, 2003, 2005). After a 4h separation, mother and litter were re-joined and maternal and offspring behaviours recorded simultaneously over 15 min, using scan sampling every 20 s (Martin and Bateson, 2007). Provisioning is measured using an established protocol (Hager and Johnstone, 2003, 2005, 2007) as maternal and offspring weight change after reunification with pups over the following two hours. Because rodents are nocturnal all observations occurred under red light, i.e. the active phase. Maternal behaviour was recorded as the sum of nursing, suckling and nest building. Nursing is defined as attending the litter, sitting on the nest and suckling up to half the litter while suckling refers to the entire litter being suckled at the same time. This distinction was used as sometimes it cannot be ascertained whether pups are suckling or not because of the position of the mothers in the nest. Nestbuilding behaviour is gathering nesting material and constructing a nest. Pup solicitation behaviour in mice is defined as pups attempting to suck and following the mother, but individual pups were not distinguished. All procedures were approved by the University of Manchester Ethics Committee.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
    Effect of within-strain sample size on QTL detection and mapping using recombinant inbred mouse strains
    1. JK Belknap
    (1998)
    Behavior Genetics 28:29–38.
  6. 6
  7. 7
  8. 8
  9. 9
    The impact of maternal uterine genotype on postnatal growth and adult body size in mice
    1. DE Cowley
    2. D Pomp
    3. WR Atchley
    4. EJ Eisen
    5. D Hawkins-Brown
    (1989)
    Genetics 122:193–203.
  10. 10
  11. 11
    Statistical methods for mapping quantitative trait loci from a dense set of markers
    1. J Dupuis
    2. D Siegmund
    (1999)
    Genetics 151:373–386.
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
    Parent–offspring conflict
    1. RM Kilner
    2. CA Hinde
    (2012)
    In: Royle NJ, Smiseth PT, Kölliker M, editors. The Evolution of Parental Care. Oxford, UK:: Oxford University Press;. 118––132.
    https://doi.org/10.1093/acprof:oso/9780199692576.003.0007
  24. 24
  25. 25
  26. 26
    Parent-offspring conflict and the genetic trade-offs shaping parental investment
    1. M Kölliker
    2. S Boos
    3. JW Wong
    4. L Röllin
    5. D Stucki
    6. S Raveh
    7. M Wu
    8. J Meunier
    (2015)
    Nature Communications, 6, 10.1038/ncomms7850.
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38

Decision letter

  1. Jonathan Flint
    Reviewing Editor; Wellcome Trust Centre for Human Genetics, United Kingdom

eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.

Thank you for submitting your work entitled "Genetic variation in offspring indirectly influences the quality of maternal behaviour in mice" for consideration by eLife. Your article has been favorably evaluated by Diethard Tautz (Senior editor) and three reviewers, one of whom is a member of our Board of Reviewing Editors.

The reviewers have discussed the reviews with one another and the Reviewing editor has drafted this decision to help you prepare a revised submission.

Summary:

This paper demonstrates that indirect genetic effects can be identified and genetically mapped. The authors adopt a cross-fostering design so that in families either the mothers or the offspring are genetically variable (from the BXD recombinant inbred panel) while the corresponding offspring or mothers show no genetic variation (i.e. are from the B6 inbred strain). The authors identify three indirect genetic effects, and three direct genetic effects. Overall this study represents a substantial and well-designed effort to address an important question.

Essential revisions:

1) Design limitations need to be acknowledged and discussed. In their design the in-utero maternal effects are perfectly confounded with offspring genotypes. This means that the results of part 1 could be interpreted as "Variation in BXD mothers genotypes causes variation in in-utero environment, which causes variation in BXD offspring behaviour later in early life, which causes variation in B6 adoptive mothers provisioning". This would effectively be IGE from BXD mothers on B6 mothers.

The results of part 2 could be interpreted as "Variation in BXD mothers genotypes causes variation in in-utero environment, which causes variation in BxD offspring weight gain and solicitation in early life" and Variation in BXD mothers genotypes independently causes variation in BXD biological mothers provisioning, in which case there would not be coadaptation between BXD offspring phenotypes and BXD maternal behaviour.

This caveat is sufficiently important that its implications should be discussed in the Discussion. As the authors note in the Discussion however, overcoming this limitation would be extremely challenging.

2) The results are only just significant. Given that they have tested multiple behaviors, and both direct and indirect effects, the chances are that a proportion of the loci are likely false positives. The authors should assess a false discovery rate and provide results corrected for the total number of tests carried out.

3) The mapping approach is relatively un-sophisticated and in particular environmental effects receive scant attention. Cage effects should be accounted for in all the analyses. In addition there may be unacknowledged environmental effects, which might be captured if they included the time when the experiment was carried out as a covariate. The fact that results for the genetic effects only just achieve significance means that there is a concern that even slight contributions from covariates may be giving rise to false positives.

4) It wasn't clear why they used per pup values. Why not fit litter size should be fitted as a covariate? They should repeat the analyses with litter fitted as a covariate.

5) The phenotypic correlation observed between BXD offspring solicitation and B6 maternal behaviour could arise from environmental effects only as BXD offspring and B6 mothers share a cage. It is not strong evidence that BXD genotypes affect B6 maternal behaviour, and can only weakly be linked to the QTLs identified as no overlap was found between DGE QTLs for offspring solicitation and IGE QTLs for maternal behaviour. Genetic correlation should be computed from the individual measurements.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Genetic variation in offspring indirectly influences the quality of maternal behaviour in mice" for further consideration at eLife. Your revised article has been favorably evaluated by Diethard Tautz (Senior editor) and three reviewers, one of whom is a member of our Board of Reviewing Editors. The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

You state that you have "used the correct litter size to calculate per pup measures, namely adoptive litter size." That a litter size effect exists does not imply that per pup values are appropriate, i.e., it does not prove that per pup values are independent of litter size. This would only be correct if you could prove that total provisioning is exactly linear in litter size; the mere significance of a (linear) litter size effect is no proof thereof.

Intuitively, one would expect that provisioning per pup decreases with litter size, simply because there is some sort of limit to total provisioning that is feasible. If you can fit a litter size effect on total provisioning using linear models, then you can also do this on provisioning defined per pup.

While this may not be feasible in the Genenetwork software that you used for the mapping, it could be investigated in a prior analysis using simple linear models. Then the results should be reported in the manuscript, and if the litter size effect on per pup provisioning is significant in the prior analysis, then you should acknowledge in the manuscript that ideally a covariate for litter size would have been included in the mapping.

The correlation/covariance that may exist between direct and indirect environmental effects (IDE and IEE) should be modelled when calculating the correlation between mother and offspring phenotypes as it may increase or decrease the correlation reported in the manuscript.

Example of such a correlation between IDE and IEE: differences in noise across cages may lead to B6 mothers in noisy cages showing reduced maternal behaviour and BXD offspring in those same cages soliciting less (or more).

To account for this covariance, you would need to focus on the genetic component of the correlation and use the replicate structure of the data (3 replicates per genotype) to calculate something analogous to a genetic correlation but where phenotypes are measured in B6 mothers and BXD offspring (instead of in the same individual as is traditional) and the genotypes would be those of the BXD offspring.

Although a sentence was added in the Discussion with regard to in-utero effects, you still need to put the results of part 1 in perspective and address interpretation of the results of part 2 (co-adaptation).

https://doi.org/10.7554/eLife.11814.017

Author response

Essential revisions:

1) Design limitations need to be acknowledged and discussed. In their design the in-utero maternal effects are perfectly confounded with offspring genotypes. This means that the results of part 1 could be interpreted as "Variation in BXD mothers genotypes causes variation in in-utero environment, which causes variation in BXD offspring behaviour later in early life, which causes variation in B6 adoptive mothers provisioning". This would effectively be IGE from BXD mothers on B6 mothers.

The results of part 2 could be interpreted as "Variation in BXD mothers genotypes causes variation in in-utero environment, which causes variation in BxD offspring weight gain and solicitation in early life" and Variation in BXD mothers genotypes independently causes variation in BXD biological mothers provisioning, in which case there would not be coadaptation between BXD offspring phenotypes and BXD maternal behaviour.

This caveat is sufficiently important that its implications should be discussed in the Discussion. As the authors note in the Discussion however, overcoming this limitation would be extremely challenging.

We agree, and have thus expanded the first paragraph of the Discussion following this suggestion though we note coadaptation may well be manifest in utero (see e.g. Wolf and Hager 2006 PLoS Biol 10.1371/journal.pbio.0040380, and 2009 in J Evol Biol 22:2519).

2) The results are only just significant. Given that they have tested multiple behaviors, and both direct and indirect effects, the chances are that a proportion of the loci are likely false positives. The authors should assess a false discovery rate and provide results corrected for the total number of tests carried out.

The experiment was designed to test the hypothesis that genetic variation in BXD offspring can affect trait variation in B6 maternal behaviour and vice versa in the context of family interactions. While we need protection against many false discoveries we need, at the same time, assurance we are not neglecting truly interesting locations (see e.g. Chen and Storey 2005 Genetics 173: 2371–2381). Traditionally, LOD scores above 3.3 are considered in interval mapping studies and we follow the convention of previous studies on maternal behaviour (e.g. Peripato et al. 2002 Genetics 162:1341) and other multiple complex traits (Cheverud et al. 2008 PNAS 105:4253, Wolf et al. 2008 PLoS Genet 4:e1000091, Hager et al. 2012 Nat Commun 3:1079) in reporting significance both at the genome-wide and suggestive level, which we exceed with five offspring and maternal traits for all loci (0.63/5 or 10,respectively).

In addition, our result that variation in maternal behaviour is affected by our identified (IGE) in offspring is further supported by the phenotypic correlation: “offspring solicitation behaviour in […] BXD pups is positively correlated with the level of maternal behaviour […] in B6 adoptive mothers on all three days we measured behaviour with significant effects of maternal behaviour and day on offspring solicitation[…]” (Results, second paragraph), notwithstanding the caveat added in the Discussion about in-utero effects.

3) The mapping approach is relatively un-sophisticated and in particular environmental effects receive scant attention. Cage effects should be accounted for in all the analyses. In addition there may be unacknowledged environmental effects, which might be captured if they included the time when the experiment was carried out as a covariate. The fact that results for the genetic effects only just achieve significance means that there is a concern that even slight contributions from covariates may be giving rise to false positives.

Please see next point.

4) It wasn't clear why they used per pup values. Why not fit litter size should be fitted as a covariate? They should repeat the analyses with litter fitted as a covariate.

In reply to points 3 and 4: First, we need to clarify the per pup measure used as it appears that this will answer several points raised. We did not record individual pup behaviour (where cage effects may appear) but rather for the entire litter and the per pup measure divides the score by litter size. This is because mothers may differ e.g. in their provisioning not because they have different genotypes but simply because they nurse litters of different sizes. Using linear models we found significant litter size effects. Thus, using a per pup measure is fully justified and appropriate, as well as supported by statistics, and we have used the correct litter size to calculate per pup measures, namely adoptive litter size. We have clarified this in the Methods section now. All our analyses are conduced within Genenetwork, the analytical tool kit for the BXD population (www.genenetwork.org) used in over 1000 papers, which at present does not allow running models with covariates but we understand this will be implemented in future.

While we unfortunately do not have the time of each recording we have experimentally ensured as much as possible to standardize motivation in mothers and offspring. This is a technique we have employed and developed over years (e.g. Hager and Johnstone 2003 Nature 421:533, 2005 Ethology 111:705, 2007 Anim Behav 74:139, Lyst et al. 2012 PLoS One 7:e47640) for the following reasons: consider the scenario of measuring nursing behaviour. Prior to the experiment on the day we have no information about whether mothers have just been nursing their pups or whether the last bout of nursing was several hours ago. Thus, we might see a difference in the pups’ motivation to suck, and in the mothers to nurse, not because of differences in genotypes but because of differences in the times that elapsed between the last feed. Therefore, we standardize the time prior to us measuring nursing.

5) The phenotypic correlation observed between BXD offspring solicitation and B6 maternal behaviour (part 1.2) could arise from environmental effects only as BXD offspring and B6 mothers share a cage. It is not strong evidence that BXD genotypes affect B6 maternal behaviour, and can only weakly be linked to the QTLs identified as no overlap was found between DGE QTLs for offspring solicitation and IGE QTLs for maternal behaviour. Genetic correlation should be computed from the individual measurements.

Cage effects may arise by a shared environment among multiple offspring that have been individually phenotyped but this is not how offspring traits were measured (see previous comment). Given that one family (i.e. one BXD genotype plus B6) occupied one cage we do not think that this effect applies here, and we have published correlations using the same protocol multiply before (e.g. Hager and Johnstone 2003 Nature 421:533, 2005 Ethology 111:705, 2006a,b Biol Lett 2:81 and 2:253). Further, our result that variation in maternal behaviour is affected by genes expressed in offspring is supported by phenotypic results: offspring solicitation behaviour in genetically variable BXD pups is positively correlated with the level of maternal behaviour in their genetically uniform B6 adoptive mothers on all three days we measured behaviour with significant effects of maternal behaviour and day on offspring solicitation. We appreciate the point the reviewer is making regarding an overlap of DGE QTL and IGE QTL, but no direct support is provided and we have no reason to assume that DGE QTL and IGE QTL have to overlap to support this argument.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Intuitively, one would expect that provisioning per pup decreases with litter size, simply because there is some sort of limit to total provisioning that is feasible. If you can fit a litter size effect on total provisioning using linear models, then you can also do this on provisioning defined per pup. While this may not be feasible in the Genenetwork software that you used for the mapping, it could be investigated in a prior analysis using simple linear models. Then the results should be reported in the manuscript, and if the litter size effect on per pup provisioning is significant in the prior analysis, then you should acknowledge in the manuscript that ideally a covariate for litter size would have been included in the mapping. We have run linear models as suggested and can confirm that litter size is a non-significant predictor for all our per pup provisioning traits.

The correlation/covariance that may exist between direct and indirect environmental effects (IDE and IEE) should be modelled when calculating the correlation between mother and offspring phenotypes as it may increase or decrease the correlation reported in the manuscript.

Example of such a correlation between IDE and IEE: differences in noise across cages may lead to B6 mothers in noisy cages showing reduced maternal behaviour and BXD offspring in those same cages soliciting less (or more). To account for this covariance, you would need to focus on the genetic component of the correlation and use the replicate structure of the data (3 replicates per genotype) to calculate something analogous to a genetic correlation but where phenotypes are measured in B6 mothers and BXD offspring (instead of in the same individual as is traditional) and the genotypes would be those of the BXD offspring.

The example given that a correlation between B6 maternal behaviour and BXD offspring solicitation could arise because all B6 shared the same environment (e.g. noise) is a very unlikely scenario as all animals are kept in the same small room (exclusively used by us), which is sound proofed, and in IVC cages, each with the same conditions. We note that the conditions experienced by B6 females are identical except the genotypes of the adoptive BXD litters, and that correlations are calculated per maternal genotype, not cage; and have added some additional information about housing to the Methods.

Although a sentence was added in the Discussion with regard to in-utero effects, you still need to put the results of part 1 in perspective and address interpretation of the results of part 2 (co-adaptation).

Yes, agreed. We have added potential implications of in-utero effects to that section in the Discussion.

https://doi.org/10.7554/eLife.11814.018

Article and author information

Author details

  1. David George Ashbrook

    Computational and Evolutionary Biology, Faculty of Life Sciences, University of Manchester, Manchester, United Kingdom
    Contribution
    DGA, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    david.ashbrook@postgrad.manchester.ac.uk
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-7397-8910
  2. Beatrice Gini

    Computational and Evolutionary Biology, Faculty of Life Sciences, University of Manchester, Manchester, United Kingdom
    Contribution
    BG, Conception and design, Acquisition of data, Analysis and interpretation of data
    Competing interests
    The authors declare that no competing interests exist.
  3. Reinmar Hager

    Computational and Evolutionary Biology, Faculty of Life Sciences, University of Manchester, Manchester, United Kingdom
    Contribution
    RH, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    Reinmar.Hager@manchester.ac.uk
    Competing interests
    The authors declare that no competing interests exist.

Funding

Natural Environment Research Council (NE/F013418/1)

  • Reinmar Hager

Biotechnology and Biological Sciences Research Council

  • David George Ashbrook

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We would like to thank Tucker Gilman, John Fitzpatrick, Casey Bergman and three anonymous reviewers for their thoughtful comments on a previous draft of the manuscript. DGA and BG are supported by the Biotechnology and Biological Sciences Research Council (BBSRC), UK and NERC grant NE/F013418/1.

Ethics

Animal experimentation: All procedures were approved by the University of Manchester Ethics Committee.

Reviewing Editor

  1. Jonathan Flint, Reviewing Editor, Wellcome Trust Centre for Human Genetics, United Kingdom

Publication history

  1. Received: September 23, 2015
  2. Accepted: December 17, 2015
  3. Accepted Manuscript published: December 23, 2015 (version 1)
  4. Version of Record published: February 9, 2016 (version 2)

Copyright

© 2015, Ashbrook et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,281
    Page views
  • 257
    Downloads
  • 3
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Comments

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry
    2. Biophysics and Structural Biology
    Lindsay D Clark et al.
    Research Article Updated